TWI377802B - Codebook searching apparatus and method thereof - Google Patents
Codebook searching apparatus and method thereof Download PDFInfo
- Publication number
- TWI377802B TWI377802B TW098127024A TW98127024A TWI377802B TW I377802 B TWI377802 B TW I377802B TW 098127024 A TW098127024 A TW 098127024A TW 98127024 A TW98127024 A TW 98127024A TW I377802 B TWI377802 B TW I377802B
- Authority
- TW
- Taiwan
- Prior art keywords
- matrices
- matrix
- precoding
- determinant
- value
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03828—Arrangements for spectral shaping; Arrangements for providing signals with specified spectral properties
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Power Engineering (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Description
1377802 P62980004TW 30636twf.doc/n 六、發明說明: 【發明所屬之技術領域】 本發明是有關於一種多輸入多輸出Multiple j叩说1377802 P62980004TW 30636twf.doc/n VI. Description of the Invention: [Technical Field of the Invention] The present invention relates to a multiple input and multiple output Multiple j叩
Multiple Output,簡稱為ΜΙΜΟ)之無線通訊系統的接收機,且 Φτ別疋有關於一種接收機的編碼薄(c〇deb〇〇k)搜尋裝置及其方 法。 【先前技術】A receiver of a wireless communication system of Multiple Output (abbreviated as ΜΙΜΟ), and Φτ does not have a codebook (c〇deb〇〇k) search device for a receiver and a method thereof. [Prior Art]
現今的通訊標準如IEEE 802.16m與LTE (Long TermToday's communication standards such as IEEE 802.16m and LTE (Long Term
Evolution)都支援閉迴路式多輸入多輸出(無線通訊傳輸技術, 其中一項很重要的技術就疋把預編碼(pre_enc〇ding)技術利用 編碼薄來實現。理論上,預編碼的預編碼成員(亦即預編碼矩 陣)越多’所需的運算複雜度也會相對地增加。目前有許多的 準則可以判斷一個預編碼矩陣是否合適,然而計算這個準則所 需要的複雜度相當大’且未必有必要將此準則完整的計算。 現今有關於編碼薄搜尋方法總共有三篇文獻,其中一篇是 北電(Nortel)在 2008 年五月時發表了 “pr〇p〇sai f〇r 802.16m Differential Encoding/Decoding for CL-MIMO Codebook Feedback”的文獻,其基本的概念與想法是出於編號 是 IEEE C802.16m-08/347 的 IEEE C802.16m Contiibution 的文 獻。另外兩篇則是德州儀器(TI)在2008年12月11日所公開的 美國專利申請案,其分別為“Reduced Search Spa^Xe&Wq^ for Codeword Selection 5 US 2008/030446广17” 與 “l〇wEvolution) supports closed-loop multi-input and multi-output (wireless communication transmission technology, one of the most important technologies is to use precoding (pre_enc〇ding) technology to achieve the use of coding thin. In theory, precoding precoding members The more (the precoding matrix) the more the required computational complexity will increase. There are many criteria to determine whether a precoding matrix is appropriate, but the complexity required to calculate this criterion is quite large 'and not necessarily It is necessary to complete the calculation of this standard. There are three papers on the codebook search method, one of which was published by Nortel in May 2008. “pr〇p〇sai f〇r 802.16m Differential The basic concepts and ideas of the Encoding/Decoding for CL-MIMO Codebook Feedback document are based on the IEEE C802.16m Contiibution document numbered IEEE C802.16m-08/347. The other two are Texas Instruments (TI). The US patent application published on December 11, 2008, respectively, is "Reduced Search Spa^Xe&Wq^ for Codeword Selection 5 US 2008/030446 Wide 17 With "l〇w
Complexity Precoding Matrix Selection,US 2008/0304464 厶1,,° 〜/ 這三篇文獻有一個共同的假設’它們都假設無線通道^ 慢變化的。因此一段時間之内,某一時間點to會先對所有編碼.二 1377802 P62980004TW 30636twf.doc/n 溥做全搜哥(full search or exhaustive search),也就是一般的作 法,接下來的數個時間點則會以前一個時間點所選出來的預編 碼矩陣做基準,做小範圍的搜尋,其中所述之範圍為已經定義 好的範圍。值得一提的是,整個程序必須每隔一段時間τ重新 整個流程,以避免產生嚴重的錯誤遞層(err〇rpr〇pagati〇n)。 假設編碼薄中有L個預編碼矩陣,每個預碼矩陣都奋去 規範離它最近的p-i ^固預編碼矩陣。所述之規範可以藉由二個 ^編碼矩_⑽來分類,_敎,代表 =,那麼下次就可以從這W個最近的預料矩_它=^ 乂個預編碼矩陣來做搜尋即可。其中,所述之分類 二 類之後必須讓傳送端與接4::==陣的分 表,因此會你用多餘的記憶體。 、、馬矩陣的關係 在一段時間T之内都僅需要從分 尋即可,而每到達整數個的時間τ時、碼矩陣做搜 搜尋。由此可見,雖然上述之編碼薄搜尋^在^做傳統的全 =減低所需搜尋的範圍,但 還=部分的時間 度的全搜尋。 間遇疋侍做最高複雜 【發明内容】 根據本示範實施例提供— :’通訊系統具有L個預編、二:的編碼薄搜尋 雜2。首先,根據L個預編碼:車的編碼溥,其令,L大於 L個矩陣為[個預編_ ^异出L個矩陣,其中, ,矩陣為方正時’執行第-步陣:接著,當L個預 -L個矩陣的對角線 :、中’弟-步驟包括 通出L個矩陣之每-個矩陣 1377802 P62980004TW 30636twf.doc/n =對角線之最大個子;找出之料㈣最大餘因子 敢小者所對應的預編碼矩陣。 *另根據本示範實施例提供一種用於通訊系統的編碼薄搜 寸方法,通訊系統具有L個預編碼矩陣的編碼薄,豆中,[大 於等於2。首先,根據L個預編碼矩陣計算出L個矩陣,立中, L個矩陣為⑽職碼輯的咐差反 編碼矩陣不為方正時,選擇性地劲许笛此平按者個預 中 了 ^雜地執m驟至第四步驟的其 r = 弟—步驟包括:計算L個矩陣的行列式;計瞀 列個矩陣之對角線的最大餘因子鱼行 則、者所對應的預編碼矩陣。其中,第二步驟包括 ^十异l個矩陣的行列式;找出L個矩陣之 對 應的預編碼矩。其+,第三步 大者所對 乘積;將L個矩陣中之對角個矩陣的對角線 留下s個矩陣,斗s 2采積較小的㈣個矩陣淘汰, 的行列式,·叶曾s· / 且大於等於1 ;計算S個矩陣 每-個矩陣的;角角線之餘因子;找出s個矩陣之 最大餘因子與行列式比==子’找出s個矩陣之對角線的 第四步驟舍括ί 者所對應的預編碼矩陣。並中, 〜驟包括.特L個矩陣 ,、π 之對角線乘積較小的積,將Μ固矩陣令 S小於L且大於等於[^渾淘太,留下S個矩陣,其中, 陣之行列式最大者所^庫tt個矩陣的行列式;找出S個矩 一、取八考所對應的預編碼矩陣。 本示範實施例提供—猶 / 置,通訊***呈有L個褚汛糸統的編碼薄搜尋裝 於2。绝戌―、有妇預編碼矩陣的編碼薄,1中 ' 、扁馬溥搜尋裝置包括 /、 L大於寺 元與決叢單元。矩陣、判別單元、計算單Complexity Precoding Matrix Selection, US 2008/0304464 厶1,, ° ~/ These three documents have a common assumption that they all assume that the wireless channel ^ is slowly changing. Therefore, within a certain period of time, at some point, the first will be all coded. 21377802 P62980004TW 30636twf.doc/n full full search or exhaustive search, which is the general practice, the next few times The point will be the benchmark of the precoding matrix selected at the previous time point, and the range is the defined range. It is worth mentioning that the entire program must be re-processed at regular intervals to avoid serious error recursions (err〇rpr〇pagati〇n). Assuming that there are L precoding matrices in the codebook, each precoding matrix strives to standardize the p-i^solid precoding matrix closest to it. The specification can be classified by two coding moments _(10), _敎, representing =, then the next time you can search from the W nearest prediction moments _ it = ^ 预 precoding matrix . Among them, the classification of the second class must be followed by the distribution of the 4::== matrix, so you will use the extra memory. The relationship between the horse matrix and the horse matrix only needs to be searched for within a period of time T, and the code matrix is searched for each time the integer time τ is reached. It can be seen that although the above-mentioned codebook search is in the range of the search required for the conventional full=decrease, it is also the full search of the partial time. The most complicated thing is to make the difference between the two. According to the exemplary embodiment, the communication system has L pre-compiled and two: encoded thin search. First, according to L precodings: the encoding code of the car, let L be larger than L matrices as [a pre-edited_^external L matrices, where, the matrix is square positively performing the first-step matrices: then, When the diagonals of L pre-L matrices: the middle-division-step includes the passage of each matrix of L matrices 1378802 P62980004TW 30636twf.doc/n = the largest of the diagonals; find out the material (4) The maximum residual factor is the precoding matrix corresponding to the small one. * Further, according to the exemplary embodiment, there is provided an encoding thin search method for a communication system having a coding thin film of L precoding matrices, in a bean, [greater than or equal to two. Firstly, L matrixes are calculated according to L precoding matrices. When L matrix is (10), the 咐 difference inverse coding matrix of the code series is not square, and the selective whistle is pre-embled. The r = brother-steps of the heterogeneous execution to the fourth step include: calculating the determinant of the L matrices; calculating the maximum cofactor of the diagonal of the matrices of the matrices, and the precoding corresponding to the fishermen matrix. Wherein, the second step comprises: a determinant of a matrix of ten different; finding a corresponding precoding moment of the L matrices. The product of +, the third step is the product of the diagonal; the diagonal of the diagonal matrix in the L matrices is left s matrices, and the s2 s 2 is mined by the smaller (four) matrices, the determinant, Ye Zeng s· / and greater than or equal to 1; calculate the matrix of each of the S matrices; the residual factor of the angular line; find the maximum cofactor of the s matrices and the determinant ratio == sub' to find s matrices The fourth step of the diagonal line encloses the precoding matrix corresponding to ί. And, the sequence includes: a special L matrix, and the product of the diagonal product of π is smaller, and the tamping matrix is such that S is less than L and greater than or equal to [^浑淘太, leaving S matrices, where, The determinant of the tl matrix of the largest determinant of the determinant; find the precoding matrix corresponding to the S moments and take the eight test. In the exemplary embodiment, the communication system is provided with a code search for L systems. Absolutely--the codebook with the maternal precoding matrix, 1 in the ', flat horses search device including /, L is greater than the temple and the continuation unit. Matrix, discriminating unit, calculation sheet
#早兀個預編碼矩陣計算出L P629S0004TW30636twf.d〇c/n 個矩陣,其t,L個矩 判別單元用以判斷L個預編=預編碼矩陣的均方差反矩陣。 算單元與決策單元^ 為” 以控制計 L個矩陣巾之對㈣* :方正時,計算單元計算 二時’計算單•個二口:;===, 餘因子與行列式,其中,j ,矩陣之對角線的最大 時’決策單元找出L個矩陣中之^ = L,編碼矩陣為方正 對應的預編碼雖,、角朗最大細子最小者所 者所對應的預編碼矩陣。、、的取大餘因子與行列式比值最小 置,、I施例提供—制於通㈣統的編㈣搜尋梦 糸統具打個預編碼矩陣的編碼簿,其中, 元轉轉單元、姻單元、計算ί 個矩陣,ΓΓ,根據個預編碼矩陣計算出L· 抱f 陣為L_編碼轉的均方差反矩陳。 ^早4以觸L_編碼轉是否為方正, 「早-與決策單個預編碼矩陣為方正時,計皙 L個矩陣中之對角線 ::凡计异 正時’計算單元計算㈣中至少為方 ^於S。當L個減碼_為妓時,決策單元拽^個中银 個預編碼矩陣不為方正時,Μ。,山應的預、扁碼 式最大者所=:=車朿早卿至少S個起陣之行列 為讓本發明之上述·和優點能更鴨紐, •丨,並配合所附圖式作詳細說明如下。 寸牛只 1377802 P62980004TW30636twf.doc/n 【實施方式】 本示範實施例提供了數種編碼 法的裝置,並選擇到合適的 又,方法及使用這些方 S前,如何選擇合適==提供給接收機使用。 中-種判別合適的預編碼F之準則的定義2有多種準則,其 二其中’ Η是-個等效的平坦衰減通道 ’ 薄中第i個預編碼矩陣,d為多輸 :J車,F[為編碼 資料數,N〇為雜訊功率,ε為 二出糸統同時所傳送的 大小為a的單位矩陣f能量’ &則表示為 繼1 咖Λ €研顺。/认)等於矩陣Ai, 貝胸k,k表不為矩陣Ai之對角線的第#早兀Precoding matrix calculates L P629S0004TW30636twf.d〇c/n matrices, and t, L moment discriminating units are used to judge the mean squared inverse matrix of L pre-coded=pre-coding matrices. The calculation unit and the decision unit ^ are "to control the pair of L matrix matrices (4) *: squared time, the calculation unit calculates the second time 'calculation list · two two ports:; ===, cofactor and determinant, where j , the maximum time of the diagonal of the matrix 'decision unit finds ^ = L in the L matrices, the encoding matrix is the precoding corresponding to the square positive, and the precoding matrix corresponding to the smallest largest subscale. , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Unit, calculate ί matrix, ΓΓ, according to a precoding matrix, calculate the mean square error inverse moment of L· 抱 f Array for L_coded. ^ Early 4 to touch L_ code to turn whether it is square, "early-and When the decision of a single precoding matrix is square, the diagonals in the L matrices are calculated:: When calculating the difference, the calculation unit calculates (4) at least squares and S. When L is reduced by _, the decision is made. The unit 拽^ a silver medium precoding matrix is not square, Μ., Shan Ying's pre- and flat code type is the largest ===朿 卿 卿 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 至少 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 137 The exemplary embodiment of the present invention provides a plurality of encoding methods, and selects the appropriate method, and how to select the appropriate == to be provided to the receiver before using the square S. The medium-type discriminates the appropriate pre-coding F The definition of the criterion 2 has a variety of criteria, the second of which is 'Η is an equivalent flat attenuation channel', the i-th pre-coding matrix in the thin, d is multi-transmission: J car, F [for the number of encoded data, N〇 is The noise power, ε is the unit matrix f energy of the size of a transmitted by the two-out system, and is expressed as 1 Λ 研 研 研 / / / / / 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于 等于For the diagonal of the matrix Ai
Ai的反矩陣,F严與hh 口值〜表不為矩陣 矩陣。 刀另J表不F(與H的厄米遜(Η—) ς事戸貝上^陣/的反矩陣Ai」為訊號的均方差(MeanThe inverse matrix of Ai, F strict and hh mouth value ~ table is not a matrix matrix. The knife is not the same as the F (with the Hermson of H (Η-), the anti-matrix Ai of the array on the mussels) is the mean square error of the signal (Mean
Sq麵Ε· ’ _為MSE)矩陣,上述的判別準則是要選到一Sq face Ε · _ _ is MSE) matrix, the above criterion is to choose one
MSE 小。因此,如果母一個預編碼矩陣都需要經過計算得到廳 ^矩陣’並且觀反_需要㈣雜高的反矩料算才能 獲得麗矩陣,因此,接收機的用以計算矩陣的晶片面積會 非常地大,而且運算時間也會非常長。附帶一提的是,上述的 準則可以適用於預編碼矩陣為非方正與方正的情況,因此,本 示範實施例會採用此準則來說明,但上述之準則並非用以限定 本發明。 由於上述的判別準則只是要比較MsE矩陣中的對角線大 小’ MSE矩陣内之非對角的元素是完全用不到的,因此,本 1377802 P62980004TW 30636twf.d〇c/a 不範貫施例利用餘因子展開式來計算MSE矩陣,以藉此節省 MSE矩陣之非對角線元素的運算量。在此假設MSE反矩陣 Ai為3x3的方正’且表示如下: «11 α\2 αη a2l α22 α23 _β31 ayi α33 ,則其對應的反矩陣Α「ι表示如下: 一 1 :11 C12MSE is small. Therefore, if the mother precoding matrix needs to be calculated to obtain the hall matrix 'and the inverse _ need (four) high inverse anti-moment calculation to obtain the MN matrix, therefore, the receiver's wafer area used to calculate the matrix will be very Large, and the computing time will be very long. Incidentally, the above criteria may be applied to the case where the precoding matrix is not square and square. Therefore, the exemplary embodiment will be described using the standard, but the above criteria are not intended to limit the present invention. Since the above discriminant criterion is only to compare the diagonal size in the MsE matrix, the non-diagonal elements in the MSE matrix are completely unusable, therefore, this 1378802 P62980004TW 30636twf.d〇c/a is not a common example. The MSE matrix is calculated using the cofactor expansion to save the computational complexity of the off-diagonal elements of the MSE matrix. It is assumed here that the MSE inverse matrix Ai is a square of '3x3' and is expressed as follows: «11 α\2 αη a2l α22 α23 _β31 ayi α33 , then its corresponding inverse matrix Α "ι is expressed as follows: a 1: 11 C12
C 13 det⑷ '21 C22 C23 "31 C22 C33 ,其中’ MSE反轉之對祕的餘因子。"、h與〜的計算 表不如1下。011=¾ V。 22 33 a23><a32 ’ C22=aii><a33-ai3Xa3i,而且 c33=auxa22-a12xa21。 反矩陣Ai1就是矩陣Ai的行列式(determinant)乘上一 餘因子的轉置運棘anspQse ,其帽置 :進二=述的判別準則僅須取得反矩陣 < 之對角線: I二子角線的值不會因為轉置運算而改變位置。另外, 上連的3x3反矩陣運算本來兩 口、需要而十异個餘因子,現在簡化到 數j i t 和I個矩陣的行列式,從餘因子的個 ^ ^ 省了咖的運算量。由於現今標準(例如, Γ . 5 2.16m)所採用的編碼薄,其預編碼輯都是么正 (unitary)矩陣。因此,去 干哥疋A 土 e ^ w、.扁碼溥的預編碼矩陣是方正時,不管 任何預編碼矩陣所計算出的 叮个& 樣,因反矩陣之行列式值都會一 尸需要夂矩陣是方正時,上述的判別準則便 而要°十异聰反轉之對角線的餘因子即可。 接者,請參照圖1與圖2,圖1是本示範實施例所提供的 1377802 P62980004TW 30636twf.d〇c/n :種編碼$解方法之流程圖,圖2是本示範實施例所提供的 種編^馬薄搜寻裝置200之系統方塊圖。編碼薄搜尋裝置2〇〇 來執行圖1所^供之編碼薄搜尋方法,且此編碼薄搜尋裝置 :用於多,輪人多輪出之無線通訊系統的接收機,假設此接收 機的編碼薄具# L個預編碼矩陣,這些L個預編碼矩陣是么 正矩陣’且接收機採用的預編碼矩陣之判別準則可以如 所述。 編碼薄搜尋裝置200包括矩陣運算單元2〇1、判 202、計算單元2〇3與決策單元2〇4。其中,矩陣運管 竊接於判別單元2〇9盘首时- τ早兀201 曾L 7早03,判別單元2〇2轉接於計 =04。 束單元2〇4 ’而計算單元2〇3職接於決策單 ^^算^观用以接收^心與^並根據 亦g n 〇,、Ss5十异出石馬巧所對應之MSE矩陣的反矩陣, ==“。判別單元202用以判別編碼_^C 13 det(4) '21 C22 C23 "31 C22 C33 , where 'the residual factor of the MSE reversal. The calculation of ", h and ~ is not as good as 1. 011=3⁄4 V. 22 33 a23><a32'' C22=aii><a33-ai3Xa3i, and c33=auxa22-a12xa21. The inverse matrix Ai1 is the determinant of the matrix Ai multiplied by a transposing factor anspQse, and its cap is set to: the criterion for the second argument is only to obtain the inverse matrix < diagonal: I two sub-angles The value of the line does not change position due to the transpose operation. In addition, the upper 3x3 inverse matrix operation originally has two, need and ten different cofactors, and is now simplified to the number j i t and the determinant of the I matrix, and the amount of computation of the coffee is saved from the cofactor of the cofactor. Due to the coding thinness used in today's standards (for example, 2. . 5 2.16m), the precoding series are all unitary matrices. Therefore, the precoding matrix of the dry 疋A soil e ^ w, . 扁 溥 is a square positive, regardless of any pre-coding matrix calculated & amp amp 因, because the determinant value of the inverse matrix will be a corpse need When the unitary matrix is a square-positive time, the above-mentioned criterion can be used to determine the residual factor of the diagonal of the ten-invariant reversal. Referring to FIG. 1 and FIG. 2, FIG. 1 is a flowchart of a method for solving a solution of 1377802 P62980004TW 30636 twf.d〇c/n provided by the exemplary embodiment, and FIG. 2 is a flowchart provided by the exemplary embodiment. A system block diagram of the horse-thin search device 200. The codebook search device 2 is configured to perform the codebook search method provided in FIG. 1, and the codebook search device is used for a receiver of a wireless communication system with multiple rounds of multiple rounds, assuming the encoding of the receiver Thin L L precoding matrices, these L precoding matrices are a positive matrix ' and the discriminant criteria of the precoding matrix employed by the receiver can be as described. The codebook search device 200 includes a matrix operation unit 2, a decision 202, a calculation unit 2〇3, and a decision unit 2〇4. Wherein, when the matrix management pipe is hacked to the discriminating unit 2〇9, the head of the disc is - τ 兀 201 has L 7 early 03, and the discriminating unit 2 〇 2 is transferred to count = 04. The beam unit 2〇4' and the computing unit 2〇3 are connected to the decision table ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ , == ". The discriminating unit 202 is used to discriminate the encoding _^
並 結果指示計算單元期根據L 個矩陣Αι〜AL計算出判別準則所需| 像 別,士杲浐干、、灰结。》— /上而要使用數據,以及根據判 不决泉早兀2〇4如何根據計算單元2〇3And the result indicates that the calculation unit period calculates the required criterion according to the L matrices 〜ι~AL | the image, the gentry, and the gray knot. 》—/I want to use the data, and according to the judgment, the spring is early 2兀4 according to the calculation unit 2〇3
,=所要採用的預編碼矩陣。決策單元2〇 J m削纖自帽…繼機所= 值得一提的是’矩陣運算單元9m Fi與等效的平坦衰減通道響應_ 可以先將預編碼矩陣 陣B(B=HFi)。接著,矩陣運算單元订矩陣相乘,以產生矩 矩陣Bh。之後,矩陣運算翠元2 叶算矩陣B的厄米遜 米遜矩陣B'd、_ ss來計算矩=:陣Bh、矩陣B的厄 -祀|早 Ai(Ai=BHB + (dN〇/Ss))。 1377802 P629S0004TW 30636twf doc/n 但疋,上述之矩陣運算單元2〇1計算 限定本發明。 T'-輯^的方式亚不用以 請繼續參照圖1與圖2,在步驟Sl〇〇 2索:二ί刀始為1。接著,在步驟S101中,編碼薄: 的矩陣運算單元201會計算矩陣A,,其中,矩陳Α‘ΐ 巧丑坪+ (代从)。接著,在步驟Sl〇爲、 !寺於 的矩陣運算單元201檢查i是否_ 置_ ϋ步ί咖,否則,則進行步驟s1G4。在二:^則進 哥裝置200的矩陣運算單元2〇1會 ·編碼 簡單地說,由步驟S10()〜S1() · σ ,/、即1=1+1。 A^Al〇 腿所組成的迴_會計算出矩陣 接著,在步驟S104中,編碼薄拇聋 202會判斷編碼薄的預編竭矩陣是否為方= ❸判別單元 搜尋裝置200的判別單元2〇2會指干若是’則編碼薄 i04進行步驟_〜sh)7,若否,則;3與決策單元 單兀204進行步驟S108〜Sill。、 D #早兀203與決策 如同前面所述,當編碼薄的 矩陣fwL所計算出矩陣Αι〜α·’^ 為方正時,預編碼 此,在步驟S105中,編巧籍指、仃列式值都會—樣。因 需要計算L個矩陣ΑΛ 置的計算單元2〇3僅 ,中’編碼薄搜尋裝置的:夕者,在步驟 說,從i等於!開始,找出角線的最大餘因子。也就是 直到i等於L為止。 +Ai之對角線的餘因子最大者, 2〇4找出\ ^驟Sl07中’編碼薄搜尋裝置20〇的決藍 找出乙個矩陣A u的决朿早兀 L之對角線的最大餘因子最小者所對 1377802 P62980004XW 30636twf.doc/n 接〜?使用。例如:矩陣物角線之最大餘因子, = the precoding matrix to be used. The decision unit 2 〇 J m is cut from the cap... the relay is = It is worth mentioning that the 'matrix operation unit 9m Fi and the equivalent flat attenuation channel response _ can be pre-coded matrix B (B = HFi). Next, the matrix operation unit multiplies the matrix to generate a matrix of matrices Bh. After that, the matrix operation Cuiyuan 2 leaves the matrix of the Hermitian Mison matrix B'd, _ ss to calculate the moment =: array Bh, matrix B er-祀 | early Ai (Ai = BHB + (dN 〇 / Ss)). 1377802 P629S0004TW 30636twf doc/n However, the matrix operation unit 2〇1 described above is limited to the present invention. The method of T'-series ^ is not necessary. Please continue to refer to Fig. 1 and Fig. 2, in step S1 〇〇 2: 2 knives start at 1. Next, in step S101, the matrix operation unit 201 of the codebook: calculates the matrix A, where the moment Α ΐ ΐ 丑 丑 坪 + ( 代 代 代 代 代. Next, in step S1, the matrix operation unit 201 of the temple is checked whether i is set or not, otherwise, step s1G4 is performed. In the second: ^, the matrix operation unit 2 〇 1 of the device 200 will be coded. Simply, the steps S10() to S1() · σ , /, that is, 1 = 1+1. The back_accounting matrix composed of A^Al 〇 leg Next, in step S104, the coded thumb squats 202 determines whether the pre-completion matrix of the codebook is square = 判别 discriminating unit 2 of the discriminating unit search means 200 If it is 'then code then i04 proceeds to step _~sh) 7, if not, then; 3 and decision unit unit 204 proceeds to steps S108 to Sill. , D #早兀203 and decision as described above, when the coded matrix fwL calculates the matrix Αι~α·'^ is square, pre-encodes this, in step S105, edits the index, 仃列式The value will be the same. Because of the need to calculate the L matrix settings of the calculation unit 2〇3 only, the medium 'code search device: the evening, in the step, from i equals! Start by finding the maximum residual factor for the corners. That is, until i is equal to L. +Ai The diagonal factor of the largest factor, 2〇4 finds out \^Sl07 in the 'codebook search device 20〇's blue to find the matrix of the matrix A u's diagonally as early as the diagonal of L The smallest residual factor is the one that is the 1377802 P62980004XW 30636twf.doc/n. use. For example: the maximum residual factor of the matrix corner
尋P 20!0的f對。角線的最大餘因子中最小者,則編瑪薄搜 == 04會告知接收機使用預編碼矩陣1V 編石碼矩陣為非方正時,則在步驟_十, ί ί00 ϋ計算單元服會計算L個矩陣 中丁工 就疋计异出det(A丨)〜det(AL)。接著,步驟sl〇9 中’編碼溥搜尋裝置的計算單A 2G3計算l個 AL之對角線的餘因子。接著 置200糾—-I 在步 中,編碼簿搜尋裝 陣之陣Αΐ〜々中之每一個矩 j對角線的袁大餘因子。也就是說,從i等於i開始,找出 I5 Ai之對角線的餘因子最大者,直到丨等於乙為止。 204 I要Φ著τ,在步驟SU1令’編碼薄搜尋裝置2〇0的決策單元 找出L個矩陣Ai〜Al之對角線的最大餘因子盥行列式比 值取小者所對紅賴碼料給接收敵用。例如 的對角線之最大餘因子與行列式比值為矩陣Αι 旱線 时與行列式比值中最小者,職碼薄搜尋裝置200 、’、朿早兀204會告知接收機使用預編碼矩陣巧。 、土在預編碼矩陣為非方正的情況下,若矩陣、〜Al之行列 式遠大於其對角線之最大餘因子時,則可以不需要計算矩陣 Αι〜AL之對角線的最大餘因子,而僅比較矩陣A:〜之行列 ,的=小,來減少運算量。因此,在此情況下,圖丨所提供之 :碼溥日搜尋方法的流程可以修正成圖3。請參照圖2與圖3, 圖3是本示範實施例所提供的另一種編碼薄搜尋方法之流程 圖。其中,步驟S300〜S308與圖1之步驟S100〜S108相同, 在此便不多贅述。 在步驟S309中,編碼薄搜尋裝置200的決策單元2〇4找 [S3 11 1377802 P62980004TW 3〇636lwf.doc/n 出L個矩陣Al〜AL之行列式最大者所對應之預編碼矩陣给接 · 收機使用。例如’矩陣Α』的行列式det(Aj)為矩陣Al〜AL之行 列式中最大者,則編碼薄搜尋裝置200的決策單元2〇4會告知 . 接收機使用預編碼矩陣Fj。 當預編碼矩陣為非方正時’計算矩陣行列式需要比較多的 時間,而且現今多數的預編碼矩陣都採用非方正的形式來將傳 ,能量集中到較少傳輸的信號上。本示範實施例所提供的編碼 4搜寻方法會對L個矩陣、〜入!^進行篩選,以留下§個矩陣 A, 〜As’(S小於L)。其中,上述的篩選方式可以採用哈連瑪德 (Hadamard)不等式來薛選出S個矩陣Αι,〜As,,而哈達瑪德不 鎌 等式表示如下: det(Ai) ^Y[axx x=l 其中’ axx表示矩陣A〖之對角線的元素。 當編碼薄的預編碼矩陣為非方正時,可以利用上述之哈達 瑪德不等式的概念來對L個矩陣進行篩選4先,計算出[ 個矩陣Ai〜Al之每—個矩_對聽乘積,接著,便可以把 ^個矩陣Ai〜AL中之對角線乘積較小的l_s個矩陣淘汰,以 鲁 碎選出S個矩陣Ai’〜as’來進行後續的計算。如此一來,便可 以減少L-S條陣之行列式與對角線餘因子的計算。 ,參照圖2與圖4’ ® 4是本示範實施例所提供的另一種 、扁碼薄搜尋方法之流程圖。其中,步驟$柳〜s術與圖1之 步驟S100〜S107相同’在此便不多費述。若編碼薄的預編碼 ,,不為方正’則在步驟S4〇8中’編碼薄搜尋裝置2〇〇的計 :單7G 203會5十异L個矩陣Ai〜Al中的每一個矩陣之對角線 乘積,換言之’會計算出L個對角線乘積。接著,在步驟S409 [S] 12 1377802 P62980004TW 30636twfdoc/n 中’編碼薄搜尋裝置200的計算單元2〇3將L個 中之對角線乘積較小的(1^)個矩_汰,僅留下8個 〜As’來進行之後的計算。 步驟S408與409主要是用來淘汰一些可能性較低的預編 2陣,以減傾丨所提供之編务簿搜尋方法必 個 之行航的運算量,雖關4所提供之編碼薄搜尋 方法比圖1必須多執行步驟S4G8與s.但計算A〜 對角線乘積的計算結果可叫後續計算^, ㈣ 式日守使用,因此並沒有多出任何額外的計算。 丁力 在步驟S4H)中,編碼薄搜尋裝置^的計 個矩陣奸〜As,的行列式。接著,步驟s4 曰 •簿搜尋裝置200的計算單元203計算s個矩睁&,〜A中之= 線的餘因子。接著,在步驟S412中,編碼薄 二 2單謂會找__Al,〜As,中之每一個矩^之= 線的最大餘因子。接著,在步驟S413中,編碼 j :決策單元戰出S個矩陣Αι,〜As,之物⑽=因2〇子〇 ”仃列式比值最小者所對應之預編碼_給接收機使用。 同理,使用哈德瑪德不等式來篩以個 式亦可以應關圖3所提供的編碼薄搜尋方法。私〜^的方 圖5是本示範實施觸提供的另—_碼驗尋村 圖。在這個示範實施例中,步驟S500〜S5鱼 /肌私 〜S410相同’在此便不多贅述。在步驟 裝置200的決策單元204找出s個矩陣A】,〜中'扁碼溥,号 者所對應之腦碼矩陣給接收機使用。1 5之仃列式最大 接著,繼續針對碼為非方正的情況 心 6,圖6是現今多數標準所定義的石馬之 對δ月麥照圖 、J的對Α?、表。圖6中 13 { S] 1377802 P62980004TW 30636twf.doc/n 所示的8x4預編碼矩陣的碼是非對稱性的,因此還是有可能要 計异到維度為4x4的MSE反矩陣之行列式。因此,可以針對 秩(rank)大於3以上之矩陣進行分解,以減少計算矩陣之行列 式的運具量。儘管MSE反矩陣的大小可能是4x4或者更大, 但是經由分解後,計算MSE反矩陣之餘因子就只需要計算到 大小為3x3的行列式。 y-y· 利用多輸人多輸出無線通訊系統之接收機已經 予 刀解運异單元來把一個大小為4χ4 計算γ陣,== =何W分解來 的大小不限定於4X4,其大小可以比4X4大7、、、而矩陣Αι 假設矩陣Ai表示如下: 4 ,則在矩陣Ai進行 表示如下: α\\ αη αη απ d2\ ^22 ^23 ^24 α3\ α32 α33 α34 戶 41 ^42 α43 α44 QR分解後,可以得到qtA=r,其中,r R-QTAt "11 r\2 rn r!4 0 h-i rn ^24 0 r32 r33 ^34 0 r,2 r43 ^44 。又因為det(QiV 說det(R)可以表^,因此,det(Ai)=det(Q Ai)=det⑻,也就是 ^et(^) = η, det rn r23 ^24 r32 hi ;42 ^43 ^44 14 1377802 P62980004TW 30636tvvf.doc/n 如此一來 , 卜』用分解來簡化大小大於3x3之矩陣八丨 丁列式運异,將可以減少搜尋預編碼矩陣的㈣及其運算 ' 古接著,請參關7,圖7是針· 4中之步驟S4i〇的詳 細〉瓜私圖。在此範例令,步驟S410藉由QR分解來將大小大 於3x3之矩陣a! 〜AS’進行簡化’以快速地計算其行列式。當Find the f pair of P 20!0. The smallest of the largest residual factors of the corner line, the code will be told to the receiver to use the precoding matrix 1V stone code matrix is not square, then the calculation unit will calculate in step _ ten, ί 00 00 In the L matrix, the work is done by det(A丨)~det(AL). Next, in step sl9, the calculation unit A 2G3 of the coded search device calculates the coefficient of the diagonal of one AL. Then set 200 corrections - I In the step, the codebook searches for the Yuan Dayu factor of each of the moments of the matrix jΑΐ~々. That is to say, starting from i equal to i, find the largest factor of the diagonal of I5 Ai until 丨 is equal to B. 204 I wants to Φ, in step SU1, the decision unit of the codebook search device 2〇0 finds the maximum residual factor of the diagonals of the L matrices Ai~Al, and the ratio of the collinearity is determined by the small one. Expected to receive enemy. For example, the maximum cofactor of the diagonal and the determinant ratio are the smallest of the matrix Αι dry line and the determinant ratio, and the job code search device 200, ', 朿 兀 204 will inform the receiver to use the precoding matrix. In the case where the precoding matrix is unsquared, if the determinant of the matrix and ~Al is much larger than the maximum residual factor of the diagonal, the maximum residual factor of the diagonal of the matrix Αι~AL may not be calculated. And only compare matrix A: ~ ranks, = small, to reduce the amount of computation. Therefore, in this case, the flow of the code day search method can be corrected to Figure 3. Referring to FIG. 2 and FIG. 3, FIG. 3 is a flow chart of another encoding thin search method provided by the exemplary embodiment. The steps S300 to S308 are the same as the steps S100 to S108 of FIG. 1, and will not be described here. In step S309, the decision unit 2〇4 of the codebook search device 200 finds [S3 11 1377802 P62980004TW 3〇636lwf.doc/n the precoding matrix corresponding to the largest determinant of the L matrices A1 to AL. Used by the receiver. For example, the determinant det(Aj) of the 'matrix 为' is the largest of the determinants of the matrix A1 to AL, and the decision unit 2〇4 of the codebook search device 200 informs. The receiver uses the precoding matrix Fj. When the precoding matrix is non-squared, it takes a lot of time to calculate the matrix determinant, and most of the precoding matrices today use non-squared forms to concentrate the energy onto the less transmitted signal. The code 4 search method provided by the exemplary embodiment filters L matrixes, ~ into !^ to leave § matrices A, ~As' (S is smaller than L). Among them, the above screening method can use the Hadamard inequality to select S matrices Αι, ~As, and the Hadamard inequality equation is expressed as follows: det(Ai) ^Y[axx x=l Where 'axx denotes the element of the diagonal of matrix A. When the coded precoding matrix is non-square, the concept of the Hadamard inequality described above can be used to filter the L matrices. First, calculate [each of the matrices Ai~Al-moment_pair product, Then, the l_s matrices with smaller diagonal products in the matrices Ai to AL can be eliminated, and the S matrices Ai'~as' can be selected to perform subsequent calculations. In this way, the calculation of the determinant and diagonal residual factor of the L-S array can be reduced. 2 and 4' ® 4 are flowcharts of another method for searching for a flat code thin film provided by the exemplary embodiment. Here, the step $柳~s is the same as the steps S100 to S107 of Fig. 1 and will not be described here. If the coded precoding is not square, then in step S4〇8, the codebook search device 2〇〇: the single 7G 203 will have a pair of 5 matrixes Ai~Al. The angular product, in other words, 'accounts for L diagonal products. Then, in step S409 [S] 12 1377802 P62980004TW 30636twfdoc/n, the calculation unit 2〇3 of the codebook search device 200 sets the (1^) moments of the diagonal products of the L to be small, leaving only The next 8 ~As' to carry out the calculations. Steps S408 and 409 are mainly used to eliminate some of the less probable pre-programmed 2 arrays, so as to reduce the amount of computation required by the book search method provided by the dip, although the codebook provided by 4 The method must perform steps S4G8 and s more than Figure 1. However, the calculation result of calculating the A~ diagonal product can be called the subsequent calculation ^, (4), and there is no extra calculation. In step S4H), Ding Li encodes the determinant of the matrix search device ^As. Next, in step s4, the calculation unit 203 of the book search device 200 calculates the residual factor of the s = 睁, 〜, A, = line. Next, in step S412, the code thin 2 is said to find the maximum residual factor of each of the moments __Al, ~As. Next, in step S413, the code j: the decision unit battles the S matrices Αι, 〜As, the object (10) = the precoding _ corresponding to the smallest 比 式 式 比 给 给 给 给 给 给 给 给 。 The use of Harder's inequality to sift through the formula can also be based on the codebook search method provided in Figure 3. The private map of Figure 5 is the other - code check village map provided by the demonstration implementation. In this exemplary embodiment, steps S500 to S5 are the same as the fish/muscle private to S410', and will not be described here. In the decision unit 204 of the step device 200, s a matrix A is found, and the medium is flat. The brain code matrix corresponding to the receiver is used by the receiver. The maximum of the 仃 仃 1 1 , 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大 最大, J's confrontation?, Table. The code of the 8x4 precoding matrix shown in Fig. 6 13 { S] 1377802 P62980004TW 30636twf.doc/n is asymmetry, so it is still possible to calculate the MSE with a dimension of 4x4. The determinant of the inverse matrix. Therefore, it is possible to decompose the matrix whose rank is greater than 3 or more. Calculate the amount of the determinant of the matrix. Although the size of the MSE inverse matrix may be 4x4 or larger, after the decomposition, the remainder factor of the MSE inverse matrix is calculated to only calculate the determinant of size 3x3. The receiver using the multi-input multi-output wireless communication system has already solved the operation of the unit to calculate a γ array with a size of 4χ4, and the size of the decomposed is not limited to 4X4, and its size can be larger than 4X4. The matrix Αι assumes that the matrix Ai is expressed as follows: 4 , then the representation in the matrix Ai is as follows: α\\ αη αη απ d2\ ^22 ^23 ^24 α3\ α32 α33 α34 Household 41 ^42 α43 α44 After QR decomposition , you can get qtA=r, where r R-QTAt "11 r\2 rn r!4 0 hi rn ^24 0 r32 r33 ^34 0 r,2 r43 ^44 . And because det (QiV says det(R ) can be expressed as ^, therefore, det(Ai) = det(Q Ai) = det(8), that is, ^et(^) = η, det rn r23 ^24 r32 hi ;42 ^43 ^44 14 1377802 P62980004TW 30636tvvf.doc/ n In this way, the use of decomposition to simplify the matrix octagonality of the size greater than 3x3 will reduce the search for the precoding matrix. Its operation 'Old Next, see switch 7, FIG. 7 is a step of needle-4 in the details of S4i〇> melon Private FIG. In this example, step S410 simplifies the matrix a!~AS' having a size larger than 3x3 by QR decomposition to quickly calculate its determinant. when
Is上^用QR分解來簡化行列式計算的概念套用到步; S108、S308 與 S510 中。 〆 ΐί驟Γ30,中’編瑪薄搜尋裝置2()()的判別單元202會 〜Α矩P Al〜AS’的大小是否大於3x3,若S個矩陣八1, -大小小於3X3,則通知編碼薄搜尋裝置200的計瞀單 ==行㈣_,若s個矩陣Αι’〜~,的大小大於= S7〇2 〇 SVC; ; s:;.;S701 ^ 3x3 ^ ^ 車1〜As的大小小於Is is applied to the step of simplifying the determinant calculation by QR decomposition; S108, S308 and S510. 〆ΐίΓ 30, the discriminating unit 202 of the 'Ma Ma search device 2()() will determine whether the size of the moment P Al~AS' is greater than 3x3, and if the S matrices are 八1, - the size is less than 3X3, then the notification The code list of the code search device 200 == line (four)_, if the size of s matrix Αι'~~, is larger than = S7〇2 〇SVC; ; s:;.;; S701 ^ 3x3 ^ ^ size of car 1~As Less than
所以編碼讀㈣置的計算單元 個矩陣Al,〜As,的行列式。 ㈢直按彳tS 在步驟S701中,因為s個矩陣Αι, 所以編碼薄搜尋裝置2〇〇的計瞀 3X3 ’ 日hTr早兀203會對S個矩陣A ,〜 H丁 QR分解。其中’計算單元2〇3包括了 qr分解運算單 二分解運算單元可以是接收機本身的职分解運曾 早:,在步驟S702中,編碼薄搜尋裝置200的計算ϊ 2挪根據分解後的結果計算s個矩陣A1,〜As,的 = 中,如何使用QR分解後的結果來計算s個矩陣Αι, )其 列式如同前面所述,在此便不再贅述。 、仃 ί S] 15 1^//802 P62980004TW30636twf.doc/n 、、,著三请回到圖4,在此將介紹如何簡化步驟S4i3之除 法運π的貝轭方式。然而,下述之簡化除法運算的實施方式亦 可以套=到步驟S1U。編碼薄搜尋裝置的決策單元2()4 · 所要,得是矩陣Al’〜,巾之對肖_最大餘因子與行列式 比值最小者’因此’編碼薄搜尋裝置遞的決策單元綱必須 使用除法運异。但事實上,這些除法運算可以利用交 避免。 土首先,編碼薄搜尋裝置200的決策單元204自i〜S中任 思挑選兩個值給P與j。接著,編碼薄搜尋裝置200的決策單 兀=〇4會比較矩陣Ap,與Aj,(p不等於之對角線的最大餘因子 · 與行列式比值,而比較cp以叫八^與cj/det(Aj,)的運算可以等 效為比較cPxdet(Aj,)與cjxdet(Ap,),其中,矩陣Ap,與A〗,的行 列式分別為det(Ap’)與det(Aj’)’且矩陣Ap,與a/之對角線的最 大餘因子分別為cp與cj。如果cPxdet(Aj,)小於cjxdet(Ap,),則 編碼薄搜尋裝置2〇〇的決策單元2〇4會將索引值χ設為p,並 更新j的值,其中’j為j與Ρ為先前未挑選過的數字。如果 cJxdet(Ap’)小於cpxdet(Aj’)’則編碼薄搜尋裝置2〇〇的決策單 元204會將索引值X設為j,並更新p的值,其中,卩為』與ρ # 為先前未挑選過的數字。重複上面的比較方式,直到p與』都 沒有任何的值可以挑選時,編碼薄搜尋裝置2〇〇的決策單元 204便可以找出矩陣Ai’〜As,中之對角線的最大餘因子與行列 式比值最小者所對應的預編碼矩陣; 接著,請參照圖8,圖8是針對圖4中之步驟S413的詳 細流程圖。在此先定義矩陣A!’〜as,中之對角線的最大餘因子 分別為(:1〜/在步驟S800中,編碼薄搜尋裝置2〇〇的決策單 元204將i與j分別言史為2與卜接著,在步驟S8〇1中,編碼 [S ] 16 P62980004TW 30636twf.doc/i t置20。的決策單元施判斷〜是否小於 叫,,)。若⑽卿)小於cJxdet(A,),則 的決策單元2G4會則师。若⑽^ 未小於dxdet(Ai ),則執行步驟S803。 在步驟S803中,.編碼薄搜尋裝置Therefore, the coded read (four) is set in the matrix of the matrix Al, ~As, determinant. (3) Pressing 彳tS directly In step S701, since s are matrix Αι, the code thin search device 2 〇〇 3X3 ′ day hTr 兀 203 will be decomposed into S matrix A, H □ QR. Wherein the calculation unit 2〇3 includes the qr decomposition operation, the single decomposition operation unit may be the job decomposition of the receiver itself. In step S702, the calculation of the codebook search device 200 is based on the decomposed result. Calculate the s matrices A1, ~As, =, how to use the QR decomposition results to calculate s matrices Αι,) The column is as described above, and will not be described here. , 仃 ί S] 15 1^//802 P62980004TW30636twf.doc/n,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, However, the embodiment of the simplified division operation described below can also be applied to step S1U. The decision unit 2()4 of the code search device is required to be the matrix Al'~, the pair of pairs of the towel_maximum cofactor and the determinant ratio is the smallest. Therefore, the decision unit of the codebook search device must use division. Different. But in fact, these divisions can be used to avoid intersections. First, the decision unit 204 of the codebook search device 200 arbitrarily selects two values from P to j from i to S. Next, the decision list 编码=〇4 of the codebook search device 200 compares the matrix Ap with Aj, (p is not equal to the maximum cofactor of the diagonal and the determinant ratio, and compares cp with eight and cj/ The operation of det(Aj,) can be equivalent to comparing cPxdet(Aj,) with cjxdet(Ap,), where the matrix Ap and A are the determinants of det(Ap') and det(Aj'), respectively. And the maximum cofactors of the matrix Ap, and the diagonal of a/ are cp and cj respectively. If cPxdet(Aj,) is smaller than cjxdet(Ap,), the decision unit 2〇4 of the codebook search device 2〇〇 will The index value is set to p, and the value of j is updated, where 'j is j and Ρ is a previously unselected number. If cJxdet(Ap') is smaller than cpxdet(Aj')', the code search device 2 The decision unit 204 sets the index value X to j and updates the value of p, where 卩 is 』 and ρ # is the previously unselected number. The above comparison is repeated until p and 』 have no value. When selected, the decision unit 204 of the codebook search device 2 can find the maximum cofactor and determinant of the diagonals in the matrix Ai'~As, The precoding matrix corresponding to the smallest value; next, please refer to FIG. 8. FIG. 8 is a detailed flowchart for step S413 in FIG. 4. Here, the maximum diagonal of the matrix A!'~as is defined. The remaining factors are (:1~/ in step S800, the decision unit 204 of the codebook search device 2 turns the history of i and j into 2 and then respectively, and in step S8〇1, encodes [S] 16 P62980004TW 30636twf.doc/it is set to 20. The decision unit is judged whether it is less than the call, and if. (10) is less than cJxdet(A,), then the decision unit 2G4 will be the division. If (10)^ is not smaller than dxdet(Ai), step S803 is performed. In step S803, the code thin search device
斷1是否等於^等於s,則執行步驟,ί S =ϋ!·8()4 °在㈣S8G4中’編贿搜尋裝置的決 兀 θ將1加1,亦即i=i+b在步驟S805中,編瑪薄 ft2。0的決策單元綱會選擇預編碼矩叫給二 二=較交叉乘積的方式,步驟⑷3的除法運算將 ,上所述,本不範實施例提供數種編碼薄搜尋方法及其裝 置’足些編碼紐尋S法可以藉崎算對肖_細子或行 式來簡化運算量。 雖然本發明已以多個示範實施例揭露如上,然其並非用以 限疋本發明,任何所屬技術領域中具有通常知識者,在不脫離 本發明之精神和範圍内,當可作些許之更動與潤飾,故本發明 之保護範圍當視後附之申請專利範圍所界定者為準。 【圖式簡單說明】 圖1是本示範實施例所提供的一種編碼薄搜尋方法之漭 程圖。 圖2是本示範實施例所提供的一種編碼薄搜尋裝置2的 之系統方塊圖。 圖3是本示範實施例所提供的另一種編竭薄搜尋方法之 1377802 P62980004TW 30636t\vf.doc/n 圖4是本示範實施例所提供的另一種編碼薄搜尋方法之 流程圖。 圖5是本示範實施例所提供的另一種編碼薄搜尋方法之 流程圖。 圖6是現今多數標準所定義的碼之大小的對照表。 圖7是針對圖4中之步驟S410的詳細流程圖。 圖8是針對圖4中之步驟S413的詳細流程圖。 【主要元件符號說明】 S100〜S111 :步驟 200 :編碼薄搜尋裝置 201 :矩陣運算單元 202 :判別單元 203 :計算單元 204 :決策單元 S300〜S309 :步驟 S400〜S413 :步驟 S500〜S511 :步驟 S700〜S703 :步驟 S800〜S805 :步驟If the break 1 is equal to ^ equal to s, then the step is executed, ί S =ϋ!·8() 4 ° In the (4) S8G4, the decision θ of the bribe search device will add 1 to 1, that is, i=i+b in step S805. In the middle, the decision unit outline of the ft2. 0 selects the precoding moment called the second two = the cross product, the division operation of the step (4) 3 will be described above, and the present embodiment provides several codebook searches. The method and its apparatus 'sufficient coded New Search S method can simplify the computation by Xiao Qi or Xia. While the invention has been described above in terms of a plurality of exemplary embodiments, the invention is not intended to be limited thereto, and may be modified by those skilled in the art without departing from the spirit and scope of the invention. The scope of protection of the present invention is defined by the scope of the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic diagram of a codebook search method provided by the exemplary embodiment. FIG. 2 is a system block diagram of an encoded thin search device 2 provided by the exemplary embodiment. FIG. 3 is a flow chart of another method for searching a thin film provided by the exemplary embodiment. FIG. 4 is a flowchart of another method for searching for a thin film provided by the exemplary embodiment. FIG. 5 is a flow chart of another codebook search method provided by the exemplary embodiment. Figure 6 is a comparison table of the sizes of codes defined by most of the standards today. FIG. 7 is a detailed flowchart for step S410 in FIG. FIG. 8 is a detailed flowchart for step S413 in FIG. [Description of Main Component Symbols] S100 to S111: Step 200: Codebook Search Device 201: Matrix Operation Unit 202: Discrimination Unit 203: Calculation Unit 204: Decision Units S300 to S309: Steps S400 to S413: Steps S500 to S511: Step S700 ~S703: Steps S800 to S805: Steps
Claims (1)
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW098127024A TWI377802B (en) | 2009-08-11 | 2009-08-11 | Codebook searching apparatus and method thereof |
US12/608,005 US8275060B2 (en) | 2009-08-11 | 2009-10-29 | Codebook searching apparatus and method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW098127024A TWI377802B (en) | 2009-08-11 | 2009-08-11 | Codebook searching apparatus and method thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
TW201106650A TW201106650A (en) | 2011-02-16 |
TWI377802B true TWI377802B (en) | 2012-11-21 |
Family
ID=43588585
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW098127024A TWI377802B (en) | 2009-08-11 | 2009-08-11 | Codebook searching apparatus and method thereof |
Country Status (2)
Country | Link |
---|---|
US (1) | US8275060B2 (en) |
TW (1) | TWI377802B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8948800B2 (en) * | 2010-03-30 | 2015-02-03 | Qualcomm Incorporated | Systems, apparatuses, and methods to facilitate coordinated scheduling in wireless communication systems |
US11956006B2 (en) * | 2022-07-01 | 2024-04-09 | Qualcomm Incorporated | Precoding perturbation to allow large array digital post distortion |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070165738A1 (en) * | 2005-10-27 | 2007-07-19 | Barriac Gwendolyn D | Method and apparatus for pre-coding for a mimo system |
KR20080073624A (en) * | 2007-02-06 | 2008-08-11 | 삼성전자주식회사 | Codebook generating method for multi-polarized mimo system and device of enabling the method |
KR101285595B1 (en) * | 2007-03-16 | 2013-07-15 | 퍼듀 리서치 파운데이션 | Aparatus for generating precoding matrix codebook for mimo system and method for the same |
US7769098B2 (en) * | 2007-06-05 | 2010-08-03 | Texas Instruments Incorporated | Low complexity precoding matrix selection |
US7751494B2 (en) * | 2007-06-05 | 2010-07-06 | Texas Instruments Incorporated | Reduced search space technique for codeword selection |
US8179775B2 (en) * | 2007-08-14 | 2012-05-15 | Texas Instruments Incorporated | Precoding matrix feedback processes, circuits and systems |
-
2009
- 2009-08-11 TW TW098127024A patent/TWI377802B/en active
- 2009-10-29 US US12/608,005 patent/US8275060B2/en active Active
Also Published As
Publication number | Publication date |
---|---|
US20110038433A1 (en) | 2011-02-17 |
TW201106650A (en) | 2011-02-16 |
US8275060B2 (en) | 2012-09-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Long MDS codes for optimal repair bandwidth | |
RU2376710C2 (en) | Improved hochwald construction of unitary matrix codebooks via eigen coordinate transformations | |
CN107431511A (en) | Method and apparatus for determining the precoder parameter in cordless communication network | |
EP3174221B1 (en) | Method and arrangement in a wireless communications system | |
EP2396938B1 (en) | Precoding codebook and feedback representation | |
CN102412842B (en) | The coded method of a kind of low density parity check code and device | |
TWI377802B (en) | Codebook searching apparatus and method thereof | |
TWI483467B (en) | Precoding method and precoder for cross-polarized antenna array | |
JP2012507919A (en) | Channel-based iterative precoder selection | |
US8121075B2 (en) | Base station and scheduling method used in base station | |
RU2528153C2 (en) | Method and system for returning channel state information | |
JP6553809B2 (en) | Method and apparatus for transmission of coding indication information and precoding matrix determination | |
KR101324395B1 (en) | Differential feedback scheme for closed-loop mimo beamforming | |
WO2010057407A1 (en) | Method and device for encoding or decoding variable code length ldpc codes and encoder and decoder | |
WO2011120440A1 (en) | Method and apparatus for retransmission and communication system thereof | |
WO2014059831A1 (en) | Method, system, and device for transmitting coding instruction information and for determining pre-coding matrix | |
JP5918352B2 (en) | Method for generating a codebook of equally spaced circular arrays and obtaining codewords from the codebook | |
JPWO2014069464A1 (en) | Error correction encoding method and error correction encoding apparatus | |
WO2024045689A1 (en) | Csi compression feedback method based on deep learning | |
JP2010252342A (en) | Multi-user scheduling method in multi-input multi-output system, and apparatus therefor | |
WO2011097881A1 (en) | Method and system for multi-level feedback of channel information | |
CN107615804A (en) | New transmission plan for equipment communication | |
CN109309517B (en) | Signal transmission method and device, computer readable storage medium, and base station | |
CN112583420B (en) | Data processing method and decoder | |
CN102263578B (en) | Precoding matrix selection method and device |